Scheduler Service Production Usage Guidelines #4249
Description
The Scheduler service write to tempfs by default and any pod churn will cause data loss. This cannot be used in production.
Prod deployment requires a persistent volume mount and need to document walkthroughs creating a volume on the main cloud providers and how to mount those
By default, Scheduler service will be writing to the containers base file system which is normally always a tempfs mount, i.e. volatile memory with no persistence. You could do a local file system mount but this breaks when running on more than one node unless you have locality constraints, but that’s not normally something you ever wanna do.
Production deployments should be a cloud provided nfs volume, pre-provisioned with terraform or w.e., or using a CSI driver